Researchers have devised a digital method to virtually unfold and read letters sealed with “letterlocking”, an old practice of securing letters before envelopes
William J. Broad / New York Times : Tweets: @laura_estill , @charlescmann , @ykomska , @nytscience , @nytscience , and @eagresta Tweets: Dr. Laura Estill / @laura_estill : Thrilled to see this news/breakthrough from @letterlocking ! https://www.nytimes.com/... You can read a great chapter (by @letterlocking) on applying knowledge of letter locking to performance (Merchant of Venice) here: https://www.brepols.net/... Charles C. Mann / @charlescmann : MIT team uses hours of computer time and X-ray scans to digitally open letters that have been folded and sewn shut for centuries: https://www.nytimes.com/... Yuliya Komska / @ykomska : People think this is relevant to centuries-old letters, but have you ever tried to open an aerogramme? https://www.nytimes.com/... @nytscience : Letterlocking was once as important for deterring snoops as encryption is to your email inbox today. Unfolding letters that were secured in this way has long posed difficulties for historians. https://www.nytimes.com/... @nytscience : A virtual-reality technique could allow historians to study letters that were “locked” by their senders, and to avoid permanently damaging them during the unfolding process. https://www.nytimes.com/... Liz Agresta / @eagresta : I've attended a bunch of the MIT Libraries' conservation department's IAP courses, including a couple on letterlocking, so this is SUPER cool to me https://www.nytimes.com/...
Context & Ripple Effects
This 2021 MIT result — X-ray scans plus computation that virtually unfold letters secured by letterlocking techniques used for centuries before envelopes — was an early proof that machines can read documents no human hand can open safely. It sits at the start of an arc the corpus keeps extending: by late 2023 a student had won a contest reading text inside a carbonized Roman scroll with an ML algorithm, and researchers went on to apply AI to the Herculaneum papyri themselves.
What changed between then and now is scope: the letterlocking work showed computation could replace physical intervention on folded paper; the scroll work showed it could replace it on charred, seemingly destroyed material — the same method family applied to progressively harder substrates.
First-order effects
- Archives and libraries holding sealed or fragile correspondence gain a way to read those items without cutting, steaming, or otherwise damaging them — the MIT team's scans and computation do the opening instead of a conservator's hands.
Second-order effects
- Institutions with unreadable holdings now have an incentive to scan intact rather than intervene physically, shifting conservation budgets toward imaging and the computational pipelines that interpret the scans.
Third-order effects
- If the pattern holds — from folded letters to the Herculaneum papyri — 'reading' becomes a computational act performed on captured data while the original object stays untouched, reordering which materials scholars consider permanently lost.
The trend: Computational imaging plus machine learning is becoming the default method for reading physically inaccessible historical documents, displacing physical restoration as the first resort.